[geeks] Slow Sol10 patching on T5140

[N. Miller]

On Oct 26, 2011, at 9:03 AM, William Enestvedt wrote:

> Jon wrote, "I wonder how much of it is IO-related."
>
>   Yeah, I thought about it being disk contention. Plus every patch goes
> in once per Zone, which drives the patching time waaay up. But even
> putting one patch on one Zone still takes longer thasn I expect. Slow
> patching of more patches on multiple Zones is a recipe for overnight
> patch runs!

I recently ran into an issue during patching some boxes with sparse zones, and
while trying to figure it out did some trussing of the patch process.

Patchadd actually stats every pkginfo file for all installed packages for each
patch it tries to apply.  IMO, live upgrade is the only way to do things now.
I went through the hoops necessary to convert from UFS to ZFS on these boxes
before applying patches just so I could use live upgrade.

The problem with using patchadd -d, though is if you have a patch succeed in
the global zone, but fail in a child zone, you are stuck for trying to get the
patch to go in the zone.  My root problem that lead to the trussing was
ultimately just that problem.  Patch installed in global but not in the child
zone.  Fortunately, patches had been installed with back out data, so I could
patchrm and re-patch.  If there had been no back out data, I'm not sure what
the solution would be.